Re: [PATCH 3/4] arm64: dts: rockchip: Add base DT for rk3528 SoC

[Krzysztof Kozlowski]

On 05/08/2024 18:22, Yao Zi wrote:
> On Mon, Aug 05, 2024 at 01:47:45PM +0200, Heiko Stübner wrote:
>> Am Montag, 5. August 2024, 13:37:11 CEST schrieb Dragan Simic:
>>> On 2024-08-05 12:59, Yao Zi wrote:
>>>> On Sun, Aug 04, 2024 at 04:05:24PM +0200, Krzysztof Kozlowski wrote:
>>>>> On 04/08/2024 15:20, Yao Zi wrote:
>>>>>>>
>>>>>>>> +		compatible = "fixed-clock";
>>>>>>>> +		#clock-cells = <0>;
>>>>>>>> +		clock-frequency = <24000000>;
>>>>>>>> +		clock-output-names = "xin24m";
>>>>>>>> +	};
>>>>>>>> +
>>>>>>>> +	gic: interrupt-controller@fed01000 {
>>>>>>>
>>>>>>> Why this all is outside of SoC?
>>>>>>
>>>>>> Just as Heiko says, device tree for all other Rockchip SoCs don't have
>>>>>> a "soc" node. I didn't know why before but just follow the style.
>>>>>>
>>>>>> If you prefer add a soc node, I am willing to.
>>>>>
>>>>> Surprising as usually we expect MMIO nodes being part of SoC to be 
>>>>> under
>>>>> soc@, but if that's Rockchip preference then fine.
>>>>>
>>>>
>>>> Okay, then I would leave it as is.
>>>>
>>>> For the fixed-clock node, I think "xin24m: clock-24m { }" is okay and
>>>> follows the new rule?
>>>
>>> I find "xin24m: clock-xin24m { }" better, because keeping the "xin24m"
>>> part in /sys listing(s), for example, can only be helpful.
>>
>> I would second that :-) . Like on a number of boards we have for example
>> 125MHz gmac clock generators ... with 2 gmacs, there are 2 of them.
>>
>> I'm not sure the preferred name accounts for that?
>>
>> Similarly we also keep the naming in the regulator node,
>> it's regulator-vcc3v3-somename {} instead of just regulator-3v3 {}.
>>
> 
> "clock-xin24m" wouldn't be more descriptive than "clock-24m" and there
> are usually only a few fixed clocks in dt, thus finding corresponding
> definition isn't a problem I think.
> 
> For the gmac case, Krzysztof, do you think something like
> "clock-125m-gmac1" is acceptable, just like what has been done for
> regulators?
>

How both above fit the generic node naming rule? You add now specific
part - purpose of the clock. The purpose is obvious from
clock-output-names or label. Anyway, not important topic to nak these
patches.

Best regards,
Krzysztof